Not that you think I have been lazy all the time 😉 I took the December a bit more relaxed and didn’t work to much on the game itself. I did some work on my own AI to learn new things… is just slow work but I also got some results. Sadly ChatGPT is far from getting ‘adult’ enabled. The full ‘porn’ activation is never coming, just from the law point of view OpenAI will not go that path. Lets see if I am wrong. As the law counts for all companies, might never open on the big AI’s. Best trick to get things a lot faster would be to run my own AI with the GPT model to create content without the ‘blocks’ the online AI has.
Good news on the other side is that I created a short code segment to generate the generation prompt with the game variables. The result was veeeeery promising! As the game has all the needed variables to generate an accurate situation description and provide hair color, hair cut, body shape, breast size, bum size, eye color, makeup setting, and more, I created 2 LoRA for Becca’s and Lana’s face to test with. Becca as ‘PC’ and Lana’s model got the in game identification ‘A15’. Using all those in the generation prompt provided rather good results and the small picture size (900×600 / 490×740 max size) the pictures get rendered in a few seconds.. proper gear and the render time should drop blow 1 second. What also means, having an AI generate the display content on the fly is very feasible.
Next, I started the programming task, let the AI write the code for the game. As I use the phone as quest control in the Supermodel Quest line and ran into certain issues, I simply gave the AI the phone locations and told it to rewrite it, how it would write it. I was very surprised when the AI pointed out several code errors and logic errors in the original GirlLife code. The rewritten phone locations worked fine after replacing the GL code. I just took the ‘Magic’ (Succubus) themes out and just went with the normal themes. I rated this as a huge success. Next I will rewrite the ‘boogie man’ location, Pregnancy and female cycle. I am most likely the guy understanding those scripts best.. the GL crew is not touching them. I want to add some functions like setting Becca pregnant without the ‘lottery’ game you have in GirlLife… will works.. after the phone script rewrite I am very confident to get the AI to code the game. Right now I use it to debug the Supermodel Quest line.. is over 2.5 million characters, a Quest line GL can only dream about).. as it is mostly non sexual, perfect Quest line to learn the ropes. So far it corrected all syntax and lazy errors, rewrite the scripts and gave it back as TXT files, which I copy pasted into the game and.. worked ^^ What it pointed out was the fact that the AI doesn’t know how to handle my phone logic. What to do in case the player picks up or rejects the call. And yes, I never coded myself the reject possibility. So I write comments over the phone handling telling what should happen in both cases. Once the comments are all in, the AI will bullet proof the phone handling. I will test all the code in the beginning bit if it runs like with the phone rewrite, I might be able to skip that and the game never needs to get debugged. Main issue.. I have to write the quest concepts very detailed for the quest handling… but is just a line more to make sure all options are covered. Things will speed up.. getting the donations in to upgrade my gear would help. The moment i have the VMEM to load the GPT5.2 model things will go to a totally new dimension.
What else… hair, hair are now easy to change.. what makes content generation by hand easier.. right now I still have to do that but gets faster and faster.
No responses yet